Fetal developmental disorders and pregnancy complications pose critical public health challenges. Yet our ability to monitor prenatal development and neonatal health has not fully entered the era of genomic medicine. Recent technological advances have allowed us to monitor global gene expression in the living human fetus or newborn, raising the possibility of creating revolutionary early diagnostics for a range of developmental disorders. Yet in order to realize this vision, we need to overcome two bioinformatics challenges. First, state-of-the-art interpretation of gene expression data requires information about how genes work together as systems. However, existing systems/pathway annotation is focused predominantly on adult disorders and is inappropriate for the analysis of fetal gene expression. Second, traditional methods for designing diagnostics from gene expression data require many examples of each sample type to be recognized. For rare diseases that cannot currently be detected in utero, however, the collection of a large number of fetal expression profiles representing each disorder is impractical. Our proposal addresses these challenges via two specific aims. First, we will develop a method for discovering new, statistically-validated gene sets from expression data. We will use this method, along with literature mining, to develop new systems biology annotation specific to this developmental stage. Second, we will develop and validate a new computational paradigm for sample classification that can identify and determine the nature of rare abnormal developmental profiles. Our methods, though motivated by the study of human development, can be applied to any medical area where existing functional annotation is inadequate or the collection of typical classification data is infeasible. This project will enable the development of new diagnostics for monitoring maternal/fetal and neonatal health, which will in turn help reduce the huge public health costs of pregnancy complications and developmental disorders. This proposal will also develop new computational methodologies for interpreting gene expression data that can be applied to other fields in medicine and biology. Thus, our work will have a broad impact on the ability to translate genomic data into clinically-relevant diagnostics and treatments across the full range of human health.